Search

In this paper we discuss methods for creating avatars that use “knowledge cards,” which are used in the EgoChat system. These avatars are agents acting in place of their creators, capable of holding conversations with other users. In the method we propose, the avatar's conversational content is structured in the form of what we call knowledge cards, which are semantically tagged text fragments; by handling both avatar statements and user statements as knowledge cards that can be handled in the same way, the person behind an avatar can easily predict the content of a conversation conducted between another user and the agent, also simplifying the task of producing content for conversational production. This is because a conversation in the form of knowledge cards is essentially a matter of combining existing text fragments. In order for a conversational agent to be used in real life, it would need to be continuously updated in its conversational content, in response to the reactions of other users, changing circumstances, and so on. The system we propose maintains a log of conversations the avatar holds with users, assisting in the task of updating the conversational contents. © 2004 Wiley Periodicals, Inc. Electron Comm Jpn Pt 2, 88(1): 32–39, 2005; Published online in Wiley InterScience (www.interscience.wiley.com). DOI 10.1002/ecjb.20144 [ABSTRACT FROM AUTHOR]